Artificial Intelligence at Purdue, Schemes and Mind Maps of Artificial Intelligence

Download Artificial Intelligence at Purdue and more Schemes and Mind Maps Artificial Intelligence in PDF only on Docsity!

AI Overview - Page 1

Artificial Intelligence at Purdue

Robert Givan

Electrical & Computer Engineering

Purdue University

AI Overview - Page 2

What is Artificial Intelligence?

Many definitions vie for attention:Psychology

Engineering

Making com-puters do thingsthe way peopledo them...

Making com-puters do thingsthat people do...(by any mech-anism)

Making com-puters do thingsthat they can’tcurrently do(that seemintelligent)

AI Overview - Page 4

Some AI History

1950’s:

AI will be really easy

1960’s:

AI will be pretty easy

1970’s:

AI is really hard

1980’s:

AI is really hard, but it sells really well!

1990’s:

We can solve small pieces of AIWe can show specific progress.

Modern AI is very different from traditional AI.

Focus on Cognitive AI

Focus onInteractiveAI

AI Overview - Page 5

Traditional AI— A Caricature

Three steps to HAL 2000:

Write down what you know in a formal logic

Code up a general purpose theorem prover

Have a conversation with it:a. Translate your comments to theoremsb. Translate your questions to logical queriesc. Translate its proofs back to natural language asanswers to the queries.

Other early approaches were equally naive...

AI Overview - Page 7

Faculty doing AI work at Purdue

• (Charlie Bouman.............................. image processing)• (Carla Brodley......... data mining for computer security)• Bob Givan ..... machine learning, planning, & reasoning• Mary Harper .......speech/language/gesture recognition• Avi Kak.......machine sensory intelligence (robot vision)• Jeff Siskind ........................needs a slide all to himself!!• Phil Swain .... AI methods to enhance teaching/learning

AI Overview - Page 8

Jeff Siskind

•^

Computational models of child language acquisition

-^

Grounding natural languagesemantics in vision

-^

Visual event perception

-^

Image segmentation

-^

Parsing images with probabilisticcontext-free grammars

Work outside AI:

•^

Whole program optimization

-^

Programming environments for worldwide distributedshared source-code repositories

AI Overview - Page 10

My Own Work in AI

•^

Reasoning:

-quickly inferring the obvious

-e.g. smarter compilers

•^

Planning:

-using reasoning and learning to plan

-compact problem representation

-handling uncertainty

•^

Learning

-planning by learning from experience

-learning for branch prediction

• learning word meanings from visual input

•^

Representation:

-class-based logic

I am interested in talking to students with

interests in any of these AI areas.

AI Overview - Page 11

Deciding What is (Obviously) True

Obvious:

Easily discovered, seen, or understood; readily perceived by the

eye or the intellect; plain; evident; apparent;

Webster’s Revised Unabridged

AI Overview - Page 13

Most Programs

Obviously Terminate

Unless there is an error:We’d like a compiler that can warn us when our programsdon’t

obviously terminate.

int factorial (int n) {

if

(n == 0)return(1)

else

return( n * factorial (1+ n))

}

AI Overview - Page 14

Planning — Deciding What to do NextInput:

•^

Your knowledge of the world

-^

Your knowledge of the likely effects of your actions

-^

Some goal or utility function

Output:

•^

A “plan”: what actions should you take?

How to find the plan? How to represent the plan?How to even represent the problem?